Jump to content

Unraid 4.7 Out of memory

Bungy

By Bungy
in User Customizations

Recommended Posts

Bungy Contributor

Bungy

Posted
Posted

I'm having an unusual issue with my unraid server running out of memory.  I have 1GB of ram.  While transferring files either through samba or ssh, the server runs out of ram and the file transfer fails.  If I stop the array, the ram is cleared back up and returns to normal.  Upon bootup there is ~600MB of ram free.

 

Is this typical behavior?  I haven't noticed this in the past before I installed an additional hard drive.  I could easily just pick up more ram, but I wanted to make sure I don't have a memory leak.  It seems that 600MB of ram is a lot to use during a file transfer.

 

I've attached my syslog file.

 

Here is my build:

ECS 945GCD-M(1.0) Intel Atom 330 Intel 945GC Micro ATX Motherboard/CPU Combo

SYBA SY-PCI40010 PCI SATA II (3.0Gb/s) 4 Port RAID 0/1/5/10 JBOD Card

1x Hitachi 2tb (parity)

2x WD 2.0 TB

1x WD 1.5 TB

1x WD 1.0 TB

1x Seagate 640 GB

 

syslog.txt

Link to comment
mbryanr Collaborator

mbryanr

Posted
Posted

Noticed your board has the Atheros 8131M LAN chip....not compatible with unRAID.

 

Search "atheros nic", "atheros 8131M"

 

Examples:

http://lime-technology.com/forum/index.php?topic=13916.msg131747#msg131747

http://lime-technology.com/forum/index.php?topic=7392.msg82034#msg82034

 

Also noticed a recent review by Raj -

http://www.greenleaf-technology.com/blogs/softwareandhardware/index.php

The test results were very clear: the ECS board's Atheros AR8131 NIC was incompatible with unRAID 4.7.  The transfers would consistently fail in some way or another - sometimes the network connection would drop out, sometimes the CRC checks would find mismatches after the transfer.  Both of these are common symptoms of an incompatible NIC.  Interestingly enough, the exact same tests run on unRAID 5.0beta10 (the latest beta available at the time of testing) showed none of the same incompatibility.  Despite repeated and redundant testing, I was not able to make the Atheros AR8131 NIC fail even once when using unRAID 5.0beta10.  This indicates that there must have been some change between unRAID 4.7 and unRAID 5.0beta10 that added support for this NIC.  The unRAID 5.0 beta release notes don't indicate any specific Atheros NIC drivers added to unRAID since 4.7, but it is possible that one of the Linux Kernal updates included one.

 

Mod edit: Fixed broken link to my blog :)

- Raj

Link to comment
Bungy Contributor

Bungy

Posted
  • Author
Posted

That is really odd as I've used this same board without a problem for a year and 2 months.  What would you recommend?  Should I buy a new nic card?  Should I try to reinstall a fresh copy of unraid on my flash drive?  Or should I upgrade to unraid 5.0b10?  I'd rather not upgrade to beta software because I'm concerned about losing any data, but I'm open to suggestions.  And thanks a lot for the reply and the links.

Link to comment
cyrnel Enthusiast

cyrnel

Posted
Posted

What error messages are you seeing? What's telling you memory is full? Is the effect linear with the size of files you copy or does it seem more like a leak?

 

I've seen people accidentally fill the ramdisk when their array wouldn't mount but otherwise don't know why Samba would be doing that. ssh copies, not so sure. Multiple places that could go wrong.

 

You probably know this... Normal Unix/Linux behavior is to use non-allocated memory for cache which can appear to exhaust memory, but it's treated as lowest priority and is invalidated as soon as something else makes a request. This behavior wouldn't cause transfers to fail but I had to mention it in case you're looking at the output from free or similar.

Link to comment
Bungy Contributor

Bungy

Posted
  • Author
Posted

I'm monitoring the memory usage using the "top" command.  I don't see any error messages, it's just as I was diagnosing the problem, I noticed the ram was depleted.  As a file transfers, the memory usage continually climbs upwards until there's about 40 megabytes of memory left.  At this point, the transfer just stops in ssh and fails in windows using samba.  I upgraded my unraid server to 2gb of ram leaving me 1.6gb of ram available after a boot up.  Just transferring a 700mb file through ssh or samba cause the ram to be eventually max out out, leading me to believe it's not just a buffer issue of the file being loaded to ram before transfer (although any size file should never have this issue).  For these reasons, it appears to be a memory leak.

Link to comment
cyrnel Enthusiast

cyrnel

Posted
Posted
Just transferring a 700mb file through ssh or samba cause the ram to be eventually max out out, leading me to believe it's not just a buffer issue of the file being loaded to ram before transfer (although any size file should never have this issue)

 

If the one transfer made memory look 700MB smaller then that's normal behavior, except the part about it stopping. What's the error at the windows side? Have you tried another client system?

 

What add-ons are you running?

Link to comment
Bungy Contributor

Bungy

Posted
  • Author
Posted

Sorry, that may not have been clear.  The one 700mb transfer made the ram about 1.5-1.6gb smaller.  On the windows side it appeared as though the file was no longer available.  This may be because the remaining 20-40mb of ram was too small for samba to work with.  The only other client system I've tried is my XBMC live (Ubuntu maverick).  This connects to the unraid box through samba shares and it will play a video file until the ram runs out and then stop.  Upon trying to replay the file, it states that it is unable to locate the requested file.

Link to comment
cyrnel Enthusiast

cyrnel

Posted
Posted

So then that wouldn't be normal behavior. :) Sounds like the network connection could be involved, as mbryanr suggested. Can we get:

 

a syslog from right after one of these fatal transfers?

ifconfig eth0

ethtool eth0

cat /proc/interrupts

 

?

 

Can you think of anything at all that changed between the setup working and not?

Link to comment
Bungy Contributor

Bungy

Posted
  • Author
Posted

I just rebooted and attached my syslog and ifconfig eth0, ethtool eth0, and interrupts file.

 

One thing that I changed is I switched my 640gb drive from being assigned to the cache drive to being in the array and I added a 80gb drive as the cache drive.  No additional packages were installed.  I don't need the additional drive at this point, so if there's a way to remove it from the array without impacting the other drives, I can try that.  This issue occurs if I assign the 80gb drive to the cache or not.

 

Also, I had a red ball next to one of my drives.  I checked the drive out and found nothing wrong with it.  I rebuilt the drive without any additional write errors.  I am getting that distinct hard drive sound that sounds like a cricket chirping when I try to access the array.

 

I know this isn't a typical diagonstic statistic, but just looking at the ram dropping, it appears I'm using about 6mb of ram/sec while transferring a file through ssh.

Logs_8-25-2011.zip

Link to comment
Bungy Contributor

Bungy

Posted
  • Author
Posted

Thanks for the responses.  I was skeptical that it was the nic since this board worked great for me for over a year.  I reinstalled unraid 4.7 on my flash drive and that seems to have solved the problem.  I'm going to give it a few days and if it remains stable, I'll change the subject title to [solved].

Link to comment
  • 2 months later...

Archived

This topic is now archived and is closed to further replies.

Go to topic listing
×
×
  • Create New...